2 –

2
Cognition, 21 (1985) 37–46
Does the autistic child have a “theory of mind”?*
SIMON BARON-COHEN
ALAN M. LESLIE
UTA FRITH
MRC Cognitive Development Unit, London
Abstract
We use a new model of metarepresentational development to predict a cognitive
deficit which could explain a crucial component of the social impairment in
childhood autism. One of the manifestations of a basic metarepresentational capacity is a ‘theory of mind’. We have reason to believe that autistic children lack
such a ‘theory’. If this were so, then they would be unable to impute beliefs to
others and to predict their behaviour. This hypothesis was tested using Wimmer
and Perner’s puppet play paradigm. Normal children and those with Down’s
syndrome were used as controls for a group of autistic children. Even though
the mental age of the autistic children was higher than that of the controls, they
alone failed to impute beliefs to others. Thus the dysfunction we have postulated and demonstrated is independent of mental retardation and specific to
autism.
1. Introduction
Childhood autism is a severe developmental disorder. It is a rare condition,
affecting about 4 in every 10,000 children. The diagnostic criteria at present are
behavioural (American Psychiatric Association, 1980; Kanner, 1943; Ritvo &
Freeman, 1978; Rutter, 1978) and the main symptom, which can be reliably
identified, is impairment in verbal and nonverbal communication. This
impairment is part of the core feature of childhood autism, namely a profound
disorder in understanding and coping with the social environment, regardless
*We are grateful to John Morton for his helpful comments on earlier drafts of this paper. We would also like
to thank staff and children of the various schools which participated in the study. The experiment was carried
out by Simon Baron-Cohen as part of his Ph.D thesis (Dept. Psychology, University of London). Reprint
requests should be addressed to: Uta Frith, MRC Cognitive Development Unit, 17 Gordon Street, London
WC1H OAH, United Kingdom.
0010-0277/85/$3.50
© Elsevier Sequoia/Printed in The Netherlands
38
S. Baron-Cohen et al.
of IQ. Additional symptoms can occur, in particular, mental retardation,
islets of ability, and ‘insistence on sameness’. Nevertheless, the pathognomonic
symptom is failure to develop normal social relationships.
Autistic children find even the immediate social environment unpredictable
and incomprehensible. They are often said in some sense to ‘treat people and
objects alike’. Wing and Gould (1979) in their epidemiological study of
severely retarded autistic children bring out the range of socially impaired
behaviour: from total withdrawal through passivity to repetitive pestering.
Lord’s (1984) review of work on peer interaction in autistic children highlights
the low level of social competence even in able autistic children, despite
improvements due to intervention. A picture of apparently intractable social
impairment emerges in the clinical follow-up studies of autism (e.g. Kanner,
1971; Kanner, Rodriguez, & Ashenden, 1972) and in the as yet rare experimental investigations (e.g. Attwood, 1984; Martini, 1980).
Although the majority of autistic children are mentally retarded (DeMyer et
al., 1974; Wing, Yeates, Brierley, & Gould 1976), and although a number of
their symptoms may be attributable to this fact (Hermelin & O’Connor,
1970) this in itself cannot be a sufficient explanation for their social impairments. First, there are autistic children with IQ’s in the normal range, and second,
mentally retarded non-autistic children, such as Down’s syndrome, are
socially competent relative to their mental age (Coggins, Carpenter, &
Owings, 1983; Gibson, 1978).
In order to explain the specific impairments of childhood autism it is necessary,
then, to consider the underlying cognitive mechanisms independent of IQ
(Frith, 1982; Hermelin & O’Connor, 1970; Rutter, 1983). So far, nobody has
had any idea of how to characterise such mechanisms in even quasi-computational terms. In this paper we put forward a suggestion which has been
derived from a new model of metarepresentational development (Leslie,
1984, to appear). This model specifies a mechanism which underlies a crucial
aspect of social skills, namely being able to conceive of mental states: that is,
knowing that other people know, want, feel, or believe things; in short, having
what Premack and Woodruff (1978) termed a ‘theory of mind’. A theory of
mind is impossible without the capacity to form ‘second-order representations’
(Dennett, 1978; Pylyshyn, 1978). According to Leslie’s model this capacity does
not appear until the second year of life. While this capacity manifests itself
eventually in a theory of mind, Leslie shows that it also accounts for the
emergence of pretend play. An absence of the capacity to form second-order
representations, then, would lead not only to a lack of theory of mind, with the
concomitant aspects of social ineptness, but also to a lack of pretend play.
Now, it is well known that autistic children, in addition to their social handicaps, also show a striking poverty of pretend play (Sigman & Ungerer, 1981;
Does the autistic child have a “theory of mind”?
39
Ungerer & Sigman, 1981; Wing, Gould, Yeates, & Brierley, 1977; Wing &
Gould, 1979). An explanation for the lack of pretend play and its curious association with the social impairments typical of autism is not obvious, and
again the notion of mental age is not helpful for this purpose. On the one
hand, even high IQ autistic children lack pretend play, and on the other hand,
severely retarded Down’s syndrome children don’t (Hill & McCune-Nicolich,
1981). However, if we suppose that autistic children lack second-order representations, then we can make sense of the association of impairments. In
order to test this hypothesis we can make the prediction that autistic children
will lack a theory of mind. It is of course possible for autistic children to have a
theory of mind and still exhibit incompetence, since social competence must
depend on a large number of factors. However, if our prediction was proved
wrong and autistic children did show evidence of employing a theory of mind,
then we could rule out a deficiency in second-order representations. Even if our
prediction was confirmed, that is, if autistic children lacked a theory of mind,
we would still have to establish that this was a specific deficit, that is, largely
independent of general mental retardation. Thus we would have to
demonstrate (a) that even those rare autistic children whose IQ’s are in the
average range should lack this ability and (b) that non-autistic but severely retarded children, such as Down’s syndrome, should possess it.
In a seminal paper, Premack and Woodruff (1978) defined theory of mind
as the ability to impute mental states to oneself and to others. The ability to
make inferences about what other people believe to be the case in a given situation allows one to predict what they will do. This is clearly a crucial component of social skills. There is growing evidence for the ability to attribute mental
states to others, and its development from the second year of life onwards
(Bretherton, McNew, & Beeghly-Smith, 1981; MacNamara, Baker, & Olson,
1976; Shantz, 1983; Shultz, Wells, & Sarda, 1980; Shultz & Cloghesy, 1981).
A convincing demonstration that an explicit theory of mind is well within the
capacity of the normal four-year-old has been given by Wimmer and Perner
(1983). These authors developed an ingenious paradigm that can be used with
very young children based on the case where the child’s own belief is different
from someone else’s belief. In order to succeed on the task the child has to be
aware that different people can have different beliefs about a situation. Hence
this case provides the strongest evidence for the capacity to conceive of mental states (Dennett, 1978). It is this paradigm that we used in the present
study.
40
S. Baron-Cohen et al.
2. Method
2.1 Subjects
Details of the subjects are shown in Table 1. The 20 autistic children had
been diagnosed according to established criteria (Rutter, 1978). In addition
there were 14 Down’s Syndrome and 27 clinically normal preschool children.
The autistic group’s mean mental age (MA) was not only higher than that of
the Down’s Syndrome group on a non-verbal scale, but also on the more
conservative measure of a verbal scale. We assumed that for the normal group
MA would roughly correspond to chronological age (CA). Therefore, their MA
was, if anything, lower than that of the handicapped groups. We selected a
high functioning subgroup of autistic children in order to enable a stringent test
of the specific deficit hypothesis to be made. Thus, the autistic group was of a
relatively high mean IQ of 82 (derived from non-verbal MA), mostly in the
average and borderline range, i.e. 70 to 108, with only one subject scoring
less than 70. The IQ’s of the Down’s Syndrome group were rather lower with
a range from 42 to 89, and an average of 64.
Table 1. Means, SDs and ranges of Chronological Age (CA) and Mental Age (MA)
in years; months
Diagnostic groups
n
Autistic
20
Down’s syndrome
Normal
14
27
CA
Nonverbal*
MA
Verbal**
Mean
11;11
9;3
5;5
SD
3;0
2;2
1;6
Range
6;1-16;6
5;4-15;9
2;8-7;5
Mean
10;11
5;11
2;11
SD
Range
4;1
6;3-17;0
0;11
4;9-8;6
0;7
Mean
4;5
_
_
SD
0;7
Range
3;5-5;9
*Leiter International Performance Scale.
**British Picture Vocabulary Test.
MA
1;8-4;0
Does the autistic child have a “theory of mind”?
41
2.2 Procedure
The procedure is illustrated in Figure 1. There were two doll protagonists,
Sally and Anne. First, we checked that the children knew which doll was
which (Naming Question). Sally first placed a marble into her basket. Then
she left the scene, and the marble was transferred by Anne and hidden in her
box. Then, when Sally returned, the experimenter asked the critical Belief
Question: “Where will Sally look for her marble?”. If the children point to the
previous location of the marble, then they pass the Belief Question by
appreciating the doll’s now false belief. If however, they point to the marble’s
current location, then they fail the question by not taking into account the
doll’s belief. These conclusions are warranted if two control questions are
answered correctly: “Where is the marble really?” (Reality Question);
“Where was the marble in the beginning?” (Memory Question).
The control questions are crucial to ensure that the child has both knowledge of the real current location of the object and an accurate memory of
Figure 1. Experimental scenario.
42
S. Baron-Cohen et al.
the previous location. There is no reason to believe that the three questions
differ from each other in terms of psycholinguistic complexity, but of course
we hypothesize that they differ in terms of conceptual complexity. The standard
scenario was repeated using a new location for the marble, so that now there
were three different locations that the child could point at (basket, box and
experimenter’s pocket). Correct responses to all three Questions for each
of the two trials were therefore different.
3. Results
All subjects passed the Naming Question. Furthermore, all subjects without
a single exception performed without any errors for both the Reality and
Memory Questions in both trials. The Belief Question for both trials was
answered consistently by each child with the sole exception of one Down’s
Syndrome child who failed trial 1 and passed trial 2. The results for Down’s
Syndrome and normal subjects were strikingly similar. 23 out of 27 normal
children, and 12 out of 14 Down’s Syndrome children passed the Belief Question on both trials (85% and 86% respectively). By contrast, 16 of the 20
autistic children (80%) failed the Belief Question on both trials. This difference between the groups was highly significant (χ2 = 25.9, df = 2, p < .001).
All 16 autistic children who failed pointed to where the marble really was,
rather than to any of the other possible locations (p = .006, Binomial Test,
one tailed). The four autistic children who passed succeeded on both trials.
Their CA ranged from 10:11 to 15:10, their non-verbal MAs were between
8:10 and 10:8, and their verbal MAs between 2:9 and 7:0. Comparison with
data in Table 1 shows that these children were fairly average on all our
available variables. There were certainly other children of equal or greater
MA and CA who gave incorrect responses.
4. Discussion
The fact that every single child taking part in the experiment correctly
answered the control questions allows us to conclude that they all knew (and
implicitly believed) that the marble was put somewhere else after Sally had
left. The critical question was, “Where will Sally look?” after she returns.
Here a group difference appeared: Autistic children answered this question
in a distinctly different way from the others. The Down’s Syndrome and
normal preschool children answered by pointing to where the marble was
put in the first place. Thus they must have appreciated that their own knowledge
Does the autistic child have a “theory of mind”?
43
of where the marble actually was and the knowledge that could be
attributed to the doll were different. That is, they predicted the doll’s
behaviour on the basis of the doll’s belief. The autistic group, on the other
hand, answered by pointing consistently to where the marble really was. They
did not merely point to a ‘wrong’ location, but rather to the actual
location of the marble. This becomes especially clear on trial 2 where the
autistic children never pointed to the box (which had been the ‘wrong’
location on trial 1), but instead to the experimenter’s pocket—that is, again to
where the marble really was. This rules out both a position preference and a
negativism explanation. Furthermore, the autistic children were not
‘contrary’ on the Reality or Memory Questions which they always answered
correctly. Clark and Rutter (1977, 1979) investigating alleged negativism in
autistic children also found no evidence of such behaviour. The failure on the
Belief Question was also not due to random pointing. Nor could it have been
due to any failure to understand and remember the demands of the task or
the narrative since these children all answered the Naming, Memory and
Reality Questions perfectly. We therefore conclude that the autistic children
did not appreciate the difference between their own and the doll’s knowledge.
Our results strongly support the hypothesis that autistic children as a group
fail to employ a theory of mind. We wish to explain this failure as an inability
to represent mental states. As a result of this the autistic subjects are unable
to impute beliefs to others and are thus at a grave disadvantage when having
to predict the behaviour of other people. There is, however, also a suggestion
of a small subgroup of autistic children who succeeded on the task and who
thus may be able to employ a theory of mind. These children who nevertheless, by definition (American Psychiatric Association, 1980; Rutter, 1978),
exhibit social impairment, would certainly deserve further study. From Leslie’s (1984) model we would predict that if they did have the capacity to form
second-order representations, then they would also show evidence of an ability
to pretend play. Furthermore, we would predict that their social impairments
would show a rather different pattern from those autistic children who fail to
use a theory of mind.
The ability we have been testing could be considered as kind of conceptual perspective-taking skill (Shantz, 1983). However, it is important to contrast the present task with traditional perceptual perspective-taking tasks,
such as ‘line of sight’ or ‘three mountains’, where a child has to indicate what
can be seen from another point of view (Hobson, 1982; Hughes & Donaldson,
1979; Piaget and Inhelder, 1956). Such perceptual perspective-taking tasks
can be solved using solely visuo-spatial skills and in no way require imputing
beliefs to others (Cox, 1980; Huttenlocher & Presson, 1979). Hobson (1984)
has recently shown that autistic children succeed on perceptual perspective-
44
S. Baron-Cohen et al.
taking tasks with doll protagonists as well as can be expected from their MA.
This finding, Hobson argued, suggests that it is very unlikely that the cognitive abilities required in taking different points of view in perceptual situations are the same as those that underlie the autistic child’s social disability.
The results of the present study would confirm this interpretation and point
towards a crucial distinction between the understanding of perceptual situations and the attribution of higher order mental states.
We conclude that the failure shown by the autistic children in our experiment constitutes a specific deficit. It cannot be attributed to the general effects of mental retardation, since the more severely retarded Down’s syndrome children performed close to ceiling on our task. Thus we have demonstrated a cognitive deficit that is largely independent of general intellectual
level and has the potential to explain both lack of pretend play and social impairment by virtue of a circumscribed cognitive failure. This finding encourages us to continue with a theoretical framework (Leslie, 1984, to appear)
which can specify the underlying connections between pretend play, theory of
mind and social skills. Deriving further testable predictions from such a model
may lead to a new approach to the cognitive dysfunction in childhood autism
(Frith, 1984).
References
American Psychiatric Association (1980). Diagnostic and statistical manual of mental disorders (3rd ed.)
Washington DC American Psychiatric Association
Attwood, A.J. (1984) Gestures of autistic children Unpublished Ph.D Thesis, University of London
Bretherton, I , McNew, S., & Beeghly-Smith, M (1981) Early person knowledge as expressed in gestural and
verbal communication When do infants acquire a ‘theory of mind”‘ In M.E. Lamb & L.R. Sherrod
(Eds ), Infant social cognition Hillsdale, NJ Erlbaum
Clark, P , & Rutter, M (1977) Compliance and resistance in the conditioning of autistic children an exploratory
study Journal of Autism and Childhood Schizophrenia, 7, 33-48
Clark, P , & Rutter, M (1979) Task difficulty and task performance in autistic children Journal of Child
Psychology and Psychiatry, 20, 271-285
Coggins, T.E., Carpenter, R.L., & Owings, N.O. (1983) Examining early intentional communication in
Down’s Syndrome and nonretarded children British Journal of Disorders in Communication, 18, 98-106
Cox, M V (1980) Visual perspective-taking in children In M V Cox ( E d . ) , Are young children egocentric?
London Batsford Academic and Educational
DeMyer, M.K., Barton, S., Alpern, G.D., Kimberlin, C , Alien, J., Yang, E., & Steele, R.(1974) The measured
intelligence of autistic children Journal of Autism and Childhood Schizophrenia, 4, 42-60
Dennett, D (1978) Beliefs about beliefs Behavioral and Brain Sciences, 4, 568-570
Frith, U. (1982) Psychological abnormalities in early childhood psychoses I n L . W i n g & J . K . Wing
(Eds.), Handbook of Psychiatry Vol 1 Cambridge Cambridge University Press
Frith, U. (1984)Anew perspective in research on autism. In A.R.A.P.I. (Eds.),Contributions a la recherche
Does the autistic child have a “theory of mind”?
45
scienttfique sur I’autisme aspects cognitifs Pans Association pour la Recherche sur 1’Autisme et les
Psychoses In-fan tiles
Gibson, D (1978) Down’s Syndrome The psychology of mongolism Cambridge Cambridge University Press
Hermelin, B., & O’Connor, N. (1970) Psychological experiments with autistic children Oxford Pergamon Press
Hill.P.M., & McCune-Nicolich, L (1981) Pretend play and patterns of cognition in Down’s Syndrome children
Child Development, 52, 217-250
Hobson, R.P.(1982) The question of childhood egocentrism the coordination of perspectives in relation to
operational thinking Journal of Child Psychology and Psychiatry, 23, 43-60
Hobson, R.P. (1984) Early childhood autism and the question of egocentrism Journal of Autism &
Developmental Disorders, 14, 85-104
Hughes, M ,& Donaldson, M (1979) The use of hiding games for studying the coordination of viewpoints Educational Review, 31, 133-140
Huttenlocher,J., & Presson, C.C. (1979) The coding and transformation of spatial information Cognitive Psychology,
11, 373-394
Kanner, L (1943) Autistic disturbances of affective contact Nervous Child, 2, 217-250
Kanner, L (1971) Follow-up study of eleven autistic children originally reported in 1943 Journal of Autism and
Childhood Schizophrenia, 1, 119-145
Kanner, L , Rodriguez, A , & Ashenden, B. (1972) How far can autistic children go in matters of social
adaptation? Journal of Autism and Childhood Schizophrenia, 2, 9-33
Leslie, A M (1984) Pretend play and representation in infancy A cognitive approach In A. R. A. P. I. (Eds.), Contributions a la recherche scientifique sur I’autisme aspects cognitifs Pans Association pour la Recherche
sur 1’Autisme et les Psychoses Infantiles
Leslie, A M (to appear) Pretense and representation in infancy Submitted for publication
Lord, C (1984) The development of peer relations in children with autism In F.J. Morrison, C. Lord & D.P.
Keating (Eds.), Advances in applied developmental psychology (pp. 165-229) New York Academic Press
MacNamara, J., Baker, E., & Olsen, C. (1976) Four year olds’ understanding of pretend, forget, and know
evidence for propositional operations Child Development, 47, 62-70
Martini, M (1980) Structures of interaction between two autistic children In T.M. Field, S. Goldberg, D. Stern
& A.M. Sostek (Eds.), High-risk infants and children: Adult and peer interactions New York: Academic
Press
Piaget, J.,& Inhelder, B. (1956) The child’s conception of space. London Routledge & Kegan Paul
Premack, D., & Woodruff, G. (1978) Does the chimpanzee have a ‘theory of mind’? Behavioral and Brain Sciences,
4, 515-526
Pylyshyn, Z.W. (1978) When is attribution of beliefs justified? Behavioral and Brain Sciences, 4, 492-593
Ritvo, E.R., & Freeman, B.J. (1978) National Society for Autistic Children definition of the syndrome of autism
Journal of the American Academy of Child Psychiatry, 17, 565-576
Rutter, M (1978) Diagnosis and definition In M. Rutter & E. Schopler (Eds.), Autism: A reappraisal of concepts of
treatment New York Plenum
Rutter, M (1983) Cognitive deficits in the pathogenesis of autism Journal of Child Psychology and Psychiatry, 24,
513-531
Shantz, C.U. (1983) Social cognition In P.H. Mussen ( E d . ) , Handbook of Child Psychology. Vol. III. J.H. Flavell
& E.M. Markman (Eds.), Cognitive Development. New York: Wiley.
Shultz,T.R., & Cloghesy, K.(1981) Development of recursive awareness of intention Developmental Psychology,
17, 456-471.
Shultz, T.R., Wells, D., & Sarda, M. (1980) The development of the ability to distinguish intended actions from
mistakes, reflexes, and passive movements British Journal of Social & Clinical Psychology, 19, 301-310
46
S. Baron-Cohen et al.
Sigman, M , & Ungerer, J. (1981) Sensorimotor skill and language comprehension in autistic children Journal of
Abnormal Child Psychology 9, 149-165.
Ungerer J.A. & Sigman, M. (1981) Symbolic play and language comprehension in autistic children
Journal of the American Academy of Child Psychiatry, 20, 318-337.
Wimmer, H., & Perner, J. (1983) Beliefs about beliefs representation and constraining function of
wrong beliefs in young children’s understanding of deception Cognition, 13, 103-128.
Wing, L., & Gould, J. (1979) Severe impairments of social interaction and associated abnormalities in
children Epidemiology and classification Journal of Autism and Developmental Disorders, 9, 1129.
Wing, L., Gould J., Yeates, S.R., & Brierley, L (1977) Symbolic play in severely mentally retarded and
in autistic children Journal of Child Psychology & Psychiatry, 18, 167-178.
Wing, L., Yeates, S.R., Brierley, L.M., & Gould, J. (1976) The prevalence of early childhood autism
Comparison of administrative and epidemiological studies Psychological Medicine, 6, 89-100
Resume
Les auteurs presentent un nouveau modele de developpement meta-cognitif pour predire le deficit cognitif qui
rendrait compte d’un composant essentiel du handicap social de l’enfant autiste. Une des manifestations d’une
capacite de base meta-cognitive est une ‘theorie de l’esprit’. Nous avons des raisons de croire que cette theorie
fait defaut chez l’enfant autiste Celui-ci serait donc incapable d’attnbuer des croyances aux autres ou de predire
leur comportement. Cette hypothese a ete testee avec le paradigme de jeu des marionettes utilise par Wimmer et
Perner. Des enfants normaux et des enfants avec trisomie 21 ont servi de groupe controle Bien que l’age mental
des enfants autistes ait ete plus eleve que deux du groupe controle, seuls les enfants autistes n’ont pu attnbuer
aux autres des croyances. Ainsi le dysfonctionnement prevu a pu etre demontre, il s’avere independant du retard
mental et specifique a l’autiste